Field Hardware (Camera Traps!)

Date: 2022-10-27

Time: 13:00 (Panama local time in 24-hour format!)

Place: Taipir room, back table

Video-conference: N/A

Social media links and hashtags:

Facilitator: George, Pen, Darin, Jason

Notetaker: Josh

Participants: Pen, Darin, George, Jason, Harold, 20+ participants

Relevant GOSH forum threads:

Session notes:

Plan: state of the art; some ideas for hacking; and more open-ended desires for camera trapping!
How can you monitor a huge forest?
- before camera traps exist, ecologists would just have to sit and observe what animals are passing by
- camera traps can save a ton of time!
- motion sensors (passive infrared) which triggers when an animal passes (when things are working well)
- usually battery is good enough to last several months, and as long as you've got a good SD card then you'll end up with lots of cool images!
- camera traps can help you monitor things over a long period of time; you can estimate number of species (or even individuals); things that happen to animals, etc.
How do we estimate abundance?
- presence of an animal, and the estimated distance, can be used in some abundance models to estimate the total abundance
- estimating the distance can be quite difficult. There are various methods, e.g. using computer vision
- what if we use two cameras, side-by-side, to give us binocular vision? using principle of parallax, two pictures taken together at the same time side-by-side can give us better estimates for animal distance
How do we build a parallax camera trap?
- we don't want to start from scratch - there are already tons of innovations (low light environment, low powered, sturdy, etc. etc.)
- infrared sensor from one camera trap is used to trigger both cameras, side-by-side (sensor for other one is blacked out)
- some minor electrical surgery needed to export signal trigger from one camera to another (via wire)
- with this system, it's not limited to this parallax setup: you can develop any special sensor that you want to use, and connect it!
- for example you could trigger an array of camera traps, if you want to monitor a large herd of animals across a space at the same time
- this could even be used to get a 3d scan of an animal! this could have interesting scientific applications.
- technical details:
  - no signal conditioning
  - open drain, so that you can attach multiple
  - currently connecting wire is exposed: could be easily damaged?
Other ideas:
- what about using a small time-controlled light, which can be triggered and then detected when it appears on the animal, in order to estimate the distance? (time-of-flight)... or
- what other sensors would be cool to connect?
- hacking an existing hardware means that the scalability problem is solved: we just need to provide specs for hacking it, and then anyone can modify a device that is already widespread
- how do we convince manufacturer to support this (i.e. not lock things down)?
- how does motion sensor work?
- essentially it sees temperature: there are 4 sensors, in a square, and if there is any differences in the intensity of the detection in each of the 4 sensors (each one has focal area of a few mm) then it determines that there is some sort of motion, and thus triggers the camera to take a photo
- how does it work with animals that don't have body heat? e.g. butterflies: they block the background heat, and thus there is a temperature differential. this is why fluttering leaves can also cause false positive (triggers when there is no bodily motion)
- can we measure eDNA at the same time?
Abundance models
- Abundance estimation of unmarked animals based on camera-trap data (Gilbert et al., 2021) <-- excellent article reviewing all sorts of abundance models!
- there are certain types of abundance metrics (e.g. local abundance vs abundance inference; relative vs absolute abundance); as well as MANY other types of ecological variables that would be of interest to ecologists (i.e. not just abundance), including behavior, distribution (where you just want to know presence/absence i.e. species identification is necessary but not distance metric nor individual identification!)
What about other monitors?
- there is passive acoustic monitoring, eDNA, LiDAR, etc.
- there have been many efforts to combine these things, but in general there is not a strong literature / framework for multi-modal biomonitoring
Hardware specs
- some camera traps require 6AA batteries and can last almost half a year
- most support burst mode (e.g. 8 in a second)
- lots of subtleties: angle of motion detection might be wider than camera angle;
- visible light illumination (not used commonly); and ultra-red illumination
How do we turn Harold's parralax camera setup into something that any ecologist can use?
- THIS IS A QUESTION WE SHOULD AIM TO ANSWER!

Relevant links:

txti